1.1Data and method
The records are NADIR's public list of question–judgment pairs, fetched once by GET on 6 October 2026 (UTC) and kept in this study's folder with a slim index of the whole list. The list holds 1,534 pairs, every one with the status "hypothesis", filed under 10 layers and three phases of the studio's research frame: erosion (1,066 pairs), collapse (230) and transition (238). Four authorship labels occur. Three of them each cover 496 pairs. The fourth, "NADIR blackbox conditional model interpretation", covers 46: 29 in the erosion phase and 17 in the collapse phase, spread over four layers, with 14 in the layer of relationships, 13 in power, 11 in fairness and 8 in work. All 46 carry the review state "source-reviewed-interpretation-unreviewed" and all set out options.
The studio describes how these pairs were made. Its notes for the sixth edition speak of 2,880 candidate excerpts from which 496 episodes were read by hand, selected and transposed into other settings, and call the results conditional model proposals with no empirically validated claim of recovery. The list's reading note, in this study's English, says that the studio reads the choices, constraints and attitudes that events reveal, that it does not take a private conversation to settle anyone's inner thoughts or intimacy, and that each card shows the fork it kept, the context lost in transposition, the conditions under which it holds or fails, and what is not yet known. The research frame's ending, the studio adds, was supplied by its research owner and is not an observed fact.
The selection rule is one sentence. This paper reads every pair in the layer "relation", which the studio labels intimacy and relationships, in the phase "erosion", whose authorship reads "NADIR blackbox conditional model interpretation". The rule gives ten pairs. The layer holds 154 pairs, 111 of them in the erosion phase; of these, 27 are episode interpretation drafts, 32 hypothetical decision exercises, 42 episode-derived interpretations, and 10 the blackbox interpretations read here. The 4 blackbox pairs in the same layer's collapse phase are not read, so that every pair read sits before the break in the studio's frame. No other NADIR paper in this series has read the relationship layer, and none of the ten pairs appears in another.
The ten are alike in form. All have the horizon "past", the time mode "episode-excerpt-review", the scenario kind "episode-transposition", the studio's seventh edition, and the episode year 2025. Each has the basis "the model's conditional interpretation", no evidence keys and no axes, and exactly two options, each with a benefit, a cost and a "when". Their settings are a trio of friends (4 pairs), a large-firm executive (2), a part-time job, a small firm, the professional world and a small club (1 each). The Korean label for the small firm is a vulgar slang word, which this study renders "Crappy small firm".
The judgments are built in a way that matters for reading them. In all ten, the judgment opens with two or three sentences of its own and then repeats, word for word, the scene, the condition and the refutation that stand elsewhere in the pair. The opening is the only part of the judgment found nowhere else: in this study's English it makes up between 18 and 35% of a judgment's words, and 356 of the 1,418 words of the ten judgments together. This paper therefore calls those first sentences the opening, and reads the condition, the refutation, the missing fact and the options as the rest of the pair.
Two cautions from the series' rules apply. A field whose wording repeats across records is the studio's frame, not evidence about any one record. Among the ten, one field is wholly frame: the note on what was lost in transposition is the same sentence in every pair. No other field shares a run of five words across three or more of the ten, so the rest of the wording is each pair's own. And a slot that the studio's format guarantees is not a finding. That every pair has a condition, a refutation, a missing fact and two options is the format. What this paper reports is what those slots say.
The records are in Korean. This study's editorial session put every string into English, sentence by sentence and with nothing added; the Korean stays the original. Every block quote is a sentence of that English, word for word, with a source line naming the pair by a short title and the field. Every count comes from a script that reads only the kept records and prints a table to the study's folder; counts of wording were made on the Korean by fixed word lists printed beside each count.
The scenes are synthetic. They were transposed from redacted excerpts, they name no real person, and this paper treats each as a made scene, not as a report of something that happened. Five lines name a real institution, an office of state, a field or a brand, and this paper does not quote them, in the body or in the appendix. Nothing here tests whether any judgment is true of real friendships or workplaces, and nothing recommends a course of action.
1.2The sign read as the end
Each of the ten pairs is a hypothesis, a model's conditional interpretation of a synthetic scene, unreviewed, and what follows reads them as such. The opening sentences of the ten judgments speak with a certainty that the scenes do not yet support. The clearest case is the first, whose question calls a private chat room for two the starting point of a crack. The scene is undramatic. Two of three old friends open a side chat and talk about the third:
Two of three friends of ten years open a separate group chat and begin to talk behind the back of the remaining one about the failures at finding a job and the frequently cancelled plans.
The judgment's opening reads that as an ending:
A room made by two of three is never a channel of consideration. The moment that room opened, the trio was already dead. Unless the room for two is closed and the problem is faced in the room for three, the time that remains is merely a process of delaying the funeral.
Read closely, the opening does three things. Its first sentence denies a motive: whatever the two say about why they opened the room, it is "never" consideration. Its second sentence moves the ending back to the first moment of the sign. Death is not what the room may lead to; it happened "the moment that room opened", and the word "already" places the reader after the fact. Its third sentence then contradicts the second without saying so. If the room is closed and the problem faced among the three, something other than a funeral follows. An ending that can be undone by facing the problem has not yet happened. The opening keeps both claims, and the stronger one comes first.
That pattern recurs. In 3 of the ten pairs the opening or the structural reading says the end has "already" come, and 7 of the ten openings reach for a figure of death, poison, a grave or a time bomb. Measured against the matching of response to the quality of information that Ansoff describes (1975), and read as graduation, the openings do the reverse: the faintest information in the scene receives the strongest reading. This study does not say the openings are wrong about the scenes. It says that they speak as if the sign were a strong one, while the scenes give only the sign.
The exchanged glance shows how small the sign can be. The scene is a single look across a table:
At a drinking session, the moment one friend earnestly opens up about their worries, the two friends sitting opposite exchange a faint glance and swallow a cough. That fleeting glance is a wordless mockery, 'there they go again', and a signal of collusion.
The scene itself already interprets: the glance "is" mockery and collusion. The opening goes further and turns the interpretation into an instruction addressed to the reader:
The moment you see the two exchanging glances, you must stop asking. Your hurt is only their drinking snack. The end of a trio begins not with a fight but with a look of contempt.
Here the record speaks in the second person and in the imperative. In 4 of the ten openings it does one or the other, or says what must be. This paper takes no side on the instruction. What it notes is the distance between the instruction and the studio's own reading note, which says that a private exchange does not settle anyone's inner thoughts. The opening reads the inside of two people's heads from a glance; the reading note, at the level of the whole list, declines to. The two stand in the same published record and are not reconciled there.
Sias and colleagues (2004) found indirection to be the usual way out of a workplace friendship, and indirect cues are the signs here too. In 5 of the ten scenes the sign is of that kind: a look, a backbiting chat, calls avoided, a corridor walked past, greetings ignored, a name moved from one line of an email to another. The records, as this paper reads them, show decline in a relationship first in indirection. They differ on what an indirect sign can carry. For Sias and colleagues the cues are part of a process people use to step away; for the openings, a single cue is the whole verdict.
The office whose reports have stopped is the clearest workplace case:
Since a tea meeting with the president last week, the team leaders who used to queue for approvals walk past the division head's office and turn toward the office of the planning division head next door. In emails on major management issues, the division head's name has quietly slipped from the recipients down to CC.
A corridor where reports have stopped is the grave of power. Do not blame the team leaders. With a hunting dog's instinct they merely smelled the rotting corpse first.
The structural reading beside it states the rule the opening applies:
In a bureaucratised organisation, the loss of power comes as a break in language. When no one asks you for information and no one seeks your stamp, you are already a dead power.
The rule is a claim about signs: the first thing to go is talk. That is close in spirit to Sias and colleagues' indirect cues, and close in its timing to the starting point of Andersson and Pearson, Tit for Tat? (1999). But the reading collapses the time between a starting point and an outcome. The corpse is already rotting; power is already dead. The scene, read on its own terms, shows a week.
The pairs set at work read their signs through broken expectations of employment. In the small firm, a salary comes two days late and, the next week, the one developer who held the whole system resigns. The opening reads the sequence this way:
Late pay is not a simple liquidity problem but an arrhythmia of the company's heart. The place the ace fled from is not an opportunity but a poisoned chalice.
In the terms of Robinson and Rousseau (1994), this is a broken obligation of employment followed by turnover. The record's opening reads a late salary as the heart of the firm failing and a resignation as the proof. It reads the breach, in other words, as already a violation, and the turnover as already its consequence. Zhao and colleagues' distinction between a breach and the feeling of violation it may produce is the gap the opening leaves out. The rest of the pair, as section 1.3 shows, puts it back.
The trio pairs add a geometry. Caplow's two against one (1956) stands in 4 of the ten scenes, which set two against one or name a room or meeting for two. The structural readings state the geometry as a law:
In a balance held by three, the moment a private channel of communication for two is opened, the remaining one is demoted to an object of observation and a potential target of expulsion.
A closed group that cannot turn its internal structural fatigue outward seeks a temporary cohesion by making its weakest or most angular member a sacrifice.
Caplow's note also stresses that small differences among members shape which coalition forms. The records' readings make the coalition the scene's fixed outcome rather than one of several that the members' differences might produce. The question these readings leave unasked is the one the third member of a triad faces earliest: whether the two have formed a coalition at all.
The pairs then ask the reader to choose a seat. Every one of the ten pairs carries a question put to the reader as "you", and 7 of these questions offer two or more roles:
Are you the ringleader who opened the room for two, the follower who was invited and nodded along, or the scapegoat who does not yet know the room exists?
The roles map Caplow's triad onto the reader, and Williams, Ostracism (2007), suggests why the third seat is the one that hurts: exclusion is detected early, as pain and threat. The position question does not say which seat the judgment was written for. In the room for two, the opening tells someone to close the room, which only the two can do; the refutation, as the next section shows, waits on the third. The reader is placed in all three seats at once.
What this section adds to the debate on weak signals is a description of how a record can read faint information strongly: by moving the ending back to the first sign, by naming the sign's meaning as if it were seen, and by putting the reader in a seat before the scene has shown who sits where. The openings are a model's conditional proposals, unreviewed; this paper does not treat them as findings about friendship or work. It treats them as one of two voices in each pair.
1.3What would have to be seen and known
The rest of each pair, still a hypothesis about a synthetic scene, reads the same sign as a beginning. Three of its parts carry that reading: the condition, which names a second sign; the refutation, which names what would overturn the judgment; and the missing fact. The options then frame the choice that remains.
The condition is a tipping point. Andersson and Pearson, Tit for Tat? (1999), locate an incivility spiral at two points, the start and the tip. The ten pairs, as this paper reads them, mark both. The scene gives the start; the condition gives a later sign that would show the start has tipped. In the room for two:
When the conversation in the room for two has turned from 'worry' to 'mockery and fatigue'
The condition assumes what the opening denied. If the talk can still "turn" from worry to mockery, then it began as worry, and the room was, at first, the channel of consideration the opening said it never was. The second sign is what the opening's verdict needs and does not have. The same holds for the glance:
When a word slips out hinting that there was already an earlier meeting between the two that left me out
The glance alone does not show a coalition. A word about an earlier meeting would. At work the tipping sign is often an act by someone with authority, not a change in the people closest to the scene:
When the store manager imposes no discipline at all on the senior who was absent without notice and berates only the one who refused to fill in
Here the second sign moves the question from colleagues to the person who sets the rules. That shift echoes Robinson and Rousseau's framing of the psychological contract as obligations between a person and an employer. The scene's cold shoulders are colleagues'; the condition asks what the manager does.
The refutation names what would overturn the judgment, and in 6 of the ten pairs it asks for proof, a showing, a confirmation or an official act:
If it is objectively proven that the friend targeted for expulsion embezzled group assets or committed an irreversible betrayal, the expulsion is justified.
If the president's withdrawals are explained by dated receipts as temporary substitute payments for urgent club expenses, the motion is dropped.
These refutations set a standard of evidence: objective proof, dated receipts. The openings set none; they read a sign. So within each pair the judgment is held to a lower standard than its own overturning. A reading that can be reached from a glance can be withdrawn only by proof. This study notes the asymmetry without judging the scenes. It matters for the weak-signals debate because it shows a record pricing errors unevenly: it is quick to read decline and slow to read its absence. The refutation in the trio's first pair is different and worth noting, because it puts the overturning in the hands of the one excluded:
If the excluded friend first confesses their isolation and asks for help, and the shared forum of the three is restored, this judgment is withdrawn.
The judgment told the two to close their room. The refutation asks the third to speak first. Read with Williams (2007), who reports that excluded people try either to restore belonging or to assert control, the refutation picks the first response and makes it the third person's task. This paper reads the refutation, in the terms of Lee's stages (1984), as naming exposure, the moment the issue is put on the table, as the way back. The refutation does not say that exposure saves the trio. It says that exposure would show the opening was wrong.
The missing fact is the most telling part. In 9 of the ten pairs it concerns what other people know, did in secret or intend; in 8 it is put as "whether":
Whether the president has already finished interviewing a successor division head through a headhunter
How much of the company's corporate overdraft limit is used up, and whether the four social insurance contributions are in arrears
In the room for two, the unknown is whether the excluded friend has already noticed; at the table, how long the two have spent on private calls. Each of these is the fact that would decide whether the sign means what the opening says. If the president has already chosen a successor, the corridor is what the opening called it; if not, the stopped reports are a mood. If the firm's credit is used up, the late salary is the failing heart; if not, it was two days. The pair names the decisive fact and then, in its opening, speaks as if it were known. In Ansoff's terms the missing fact is the information whose quality the response should be matched to, and the records themselves say it is absent.
The options, finally, do not choose between ending a relationship and keeping it. In 6 of the ten pairs one option faces the matter directly, and in 7 one keeps a distance, withdraws or stays quiet. The room for two shows both:
Facing the discomfort directly
When the other side has the will to recover and a balance of trust remains that could clear up the misunderstanding
Gradual distancing and standing by
One becomes an accomplice to betrayal oneself, and the friendship of the three is degraded into hypocrisy for good
The first option's "when" assumes what the opening ruled out: a relationship in which trust and the sign of distrust stand together. That is the relationship Lewicki, McAllister and Bies (1998) describe, with trust and distrust held at once, and the option is written for it; the opening is written for one in which a single sign of distrust has ended trust. The second option resembles what Sias and colleagues (2004) found people at work mostly did, stepping back by indirect means; the record attaches to it a cost stated in the language of the opening, a friendship "degraded into hypocrisy for good". The study does not weigh the options. The records call them a model's conditional proposals, and the choice between them is not this paper's to make.
The workplace pairs follow the same shape with a different vocabulary. In the stopped reports, the first option is a one-on-one with the president to settle the matter, and its "when" turns on whether the president has been misled by another faction, that is, on whether the sign means what the opening said. In the disclaimer, set among auditors, the opening speaks of a time bomb, while the refutation would accept a different opinion only after an outside, independent re-appraisal:
If the reality of the inventory is publicly proven through a re-appraisal by an outside independent appraisal body, changing the opinion is warranted.
That scene is a transposed, synthetic one, and this paper reads nothing in it about any real firm or audit. Its interest here is formal. Even where the opening is most certain, the pair's own overturning waits on a public showing of facts that the scene does not contain.
What this section adds to the debate on how relationships come apart is that the pairs, taken whole, lay out the points that Baxter, Lee and Andersson and Pearson describe: a starting sign, a tipping sign, a stage of exposure or of withdrawal, and an outcome that waits on facts not yet known. Baxter, Trajectories of Relationship Disengagement (1984), found that relationships do not dissolve through a single set of stages. The pairs do not claim a single set either. Each lays out its own short course, and each opening then speaks as if the course were over.